Computer Communications 23 (2000) 17631776

www.elsevier.com/locate/comcom

Performance analysis of a xed local anchor scheme for supporting UPT services
E.S. Bae a,*, M.Y. Chung b, D.K. Sung c
Korea Telecom Telecommunications Network Lab, Seoul, 137-792, South Korea Electronics Telecommunications Research Institute, Taejon, 305-350, South Korea c Department of EE, Korea Advanced Institute of Science and Technology, Taejon, 305-701, South Korea
b a

Received 6 August 1999; revised 28 March 2000; accepted 28 March 2000

Abstract Intelligent Networks (INs) are well suited for implementing Universal Personal Telecommunication (UPT) services. For efcient planning of IN accommodating UPT services, it is essential to analyze the effect of UPT user personal mobility. In this paper, an IN architecture with a xed local anchor (FLA) is proposed for supporting UPT services. The performance of the proposed IN architecture is compared with an IN architecture based on IS-41 with consideration of UPT user personal mobility in terms of call delivery cost and location update cost. The performance of the proposed IN depends on the characteristics of UPT user mobility and incall deregistration. The proposed IN yields better performance than an IN based on IS-41 as UPT user mobility decreases and the probability of explicit incall deregistration increases. q 2000 Elsevier Science B.V. All rights reserved.
Keywords: Fixed local anchor scheme; UPT service

1. Introduction A Universal Personal Telecommunication (UPT) service will enable users to originate and receive calls at any terminal through use of personal identication numbers. This type of system is destined to become popular in the near future [1]. ITU-T has recommended network capabilities to support UPT services [2], service procedure for UPT [3], and service descriptions and operational provisions for UPT service sets 1 [4]. UPT networks must track the location of UPT users in order to support PM. PM refers to the capability of end users to originate and receive calls and to access subscribed telecommunication services on any terminal in any location, and the capability of the network to identify end users as they move. PM can be supported through a

* Corresponding author. Tel.: 182-2-526-5275; fax: 182-2-526-6970. E-mail address: esbae@kt.co.kr (E.S. Bae). Abbreviations: CCF, call control function; CRR, call to registration ratio; FLA, xed local anchor; IN, intelligent network; LSTP, local signal transfer point; PM, personal mobility; RA, registration area; RMR, registration to mobility ratio; RSTP, regional signal transfer point; SCF, service control function; SCP(H/V), service control point (home/visited); SDF, service data function; SRF, specialized resource function; SSF, service switching function; SSP, service switching point; UPT, universal personal telecommunication

unique personal number, i.e. the UPT number, in wireline and wireless networks. UPT networks have been studied in several research areas such as UPT functional architecture [59], information localization scheme [1011], service control scheme [12 14] and performance evaluation [1517]. Lauer [13] proposed local, remote, and home methods to control UPT services in wired networks and also compared these methods in terms of new service deployment, service provider requirements, and network performance. Folkestad [14] presented three models for distributing UPT service SSP and SCP in signaling networks and evaluated their performance by calculating the mean call setup time. Kwiatkowski [15] calculated the network load required to access SCPs or databases without considering UPT user mobility. Chung and Sung [16] proposed a request-based scheme for incall registration/deregistration of UPT users and incall registration resets of terminal owners, and evaluated the performance of the scheme in terms of total cost and the number of terminal prole accesses per unit time for a terminal. Chung et al. [17] also proposed an alternative scheme, i.e. a timer-based scheme for managing information related to incall registration and compared the performance of the timer-based scheme with the request-based scheme in terms of the number of signaling messages transferred between SSP and SCP per incall request reset for a terminal.

0140-3664/00/$ - see front matter q 2000 Elsevier Science B.V. All rights reserved. PII: S 0140-366 4(00)00217-6

1764

E.S. Bae et al. / Computer Communications 23 (2000) 17631776

Fig. 1. An IN architecture with FLA.

In an UPT environment, user location information is generally updated by UPT users requests, and then, the request messages are required. The characteristics of UPT user incall registration and deregistration should be considered in the design of INs for supporting UPT services. Signaling message transfers between SSPs and SCPs are required for tracking the UPT user location whenever a UPT user registers/deregisters an incall. This may cause a higher trafc load in signaling networks if the current location of a UPT user is far away from the associated SCP, and if the UPT user registers/deregisters incall frequently. The SCP involved in processing location registration/deregistration and call delivery may experience excessively high database access trafc. An IN architecture using a FLA scheme is proposed in order to support UPT services. The FLA scheme is modied from the LA scheme proposed by Ho and Akyildiz [18]. The performance of the proposed IN architecture is evaluated in terms of the call delivery cost and the location update cost. The rest of this paper is organized as follows: Section 2 briey explains some UPT service features; Section 3 proposes an IN architecture with FLA and explains incall

registration/deregistration scenarios; Section 4 derives both the call delivery cost and the location update cost; Section 5 compares the performance of the proposed IN architecture with the IS-41-based [19] IN architecture and Section 6 presents conclusions. 2. Features of UPT Service The essential features, including the following, have been dened for the operations of UPT service set 1 [4]: UPT user identity authentication is required to verify the identity of each UPT user. Incall registration enables UPT users to register for incoming calls on the current terminal. Outgoing UPT calls allow UPT users to initiate an outgoing UPT call attempt from any terminal. Incall delivery provides voice connection to UPT users who registered incalls using the registered terminal address of the UPT user regardless of the location and type of the terminal.

E.S. Bae et al. / Computer Communications 23 (2000) 17631776

1765

Fig. 2. Location of UPT user.

Incall registration provides a means for a UPT user to indicate where incoming UPT calls are to be delivered. The characteristics of a UPT users' incall registration are closely related to the PM of the UPT users. Incall registration can be cancelled by the following ways [4]: UPT users can initiate incall registration at another terminal. UPT users can explicitly request incall deregistration. UPT users can specify timer or counter values deregistering incall registration. UPT service providers can specically deregister a UPT user. Terminal owners (UPT or non-UPT) may reset all incall registrations on their terminal addresses. For simplicity, the rst two methods are here considered for incall deregistration. 3. IN architecture with xed local anchor The IN functional entities, which are required to support UPT services include a CCF, a SSF, a SCF, a SDF, and a SRF. The CCF provides call/connection processing and control. The SSF associated with the CCF provides functions required for interactions between the CCF and the SCF. The SSF also extends the logic of the CCF to include
Table 1 Six possible combinations of location updates Case User's location at the time instant of the current registration request Home Home Home Local Local Local User's location at the time instant of the next registration request Home Local Remote Home Local Remote

1 2 3 4 5 6

the recognition of service control triggers and allows interaction with the SCF. The SCF controls the CCF for the processing of UPT service requests and interacts with other functional entities to access additional logic or to obtain information (service or user location information) required to process call/service logic instances. The SCF also contains logic and processing capabilities, which are required to handle UPT service attempts. The SDF contains UPT user and network data for a real-time access by the SCF in the execution of UPT services. The SRF cooperates with the SCF to transfer information from a UPT user. The SSP can detect UPT services and communicates with other physical entities, including the SCF. In order to recognize UPT services and to control UPT calls, the SSP should include SSF, CCF and SRF functions. The SCP includes SCF and SDF functions. The SCF includes the UPT service logic program for control of UPT services. UPT user service proles, which reside in the SDF, contain all information related to UPT users, such as user-dened service information and user location information. This information provides users with UPT services. Each UPT service prole is associated with a single UPT number. Fig. 1 shows an example of intelligent network architecture consisting of N SSPs, N SCP(V)s, and one SCP(H). Each SSP is connected to the corresponding SCP(V) through a direct link and any SSP can access the data in the SCP(H) through signaling networks. The SCP(H) stores user proles and each SCP(V) stores replications of the user proles of subscribers residing in its associated SSP. Neighboring SCP(V)s connected to each LSTP are grouped. All RAs covered by an LSTP are called the LSTP region. For each UPT user, a specic SCP(V) in each LSTP region is designated as an FLA. An SSP sends an incall registration message to the FLA whenever UPT users move to a new RA in an LSTP region and register incalls on terminals located in the new RA. When the UPT users move to another LSTP region and register incalls on terminals in the new region, an SSP sends an incall registration message to the SCP(H) instead of to the FLA. Then, the SCP(H) transfers the registered user's service prole to a new FLA and sends a message to the previous FLA to delete the service prole. In order to reduce signaling load, more than one FLA may be designated. For example, in the case of one LSTP region with 10 SCP(V)s and two FLAs, two SCP(V)s are designated as FLA1 and FLA2. If a UPT user with an even or odd UPT number requests incall registration in a new RA, an incall registration message is sent to either FLA1(even) or FLA2(odd). In the architecture with a single FLA per LSTP region, the location of a UPT user can be represented using home, local and remote RAs, as shown in Fig. 2. The home RA has an FLA in an LSTP region, local RAs are all RAs except the home RA in the LSTP region, and remote RAs are located outside of the LSTP region. Table 1 shows six possible combinations of location updates according to current and next locations of a UPT

1766

E.S. Bae et al. / Computer Communications 23 (2000) 17631776

Fig. 3. Incall registration scenario in Cases 1 and 2.

user. Both the incall registration scenario and the incall registration cost are different in each case. In Case 1, when a UPT user requests incall registration for a visited SSP, the visited SSP receives an incall registration request message, then sends the messages to the corresponding SCP(V) through a direct link because the SCP(V) is designated as the UPT user's FLA. The subsequent signaling messages required for processing the incall registration are transferred between the visited SSP and the SCP(V). Incall registration overrides the previous registration. In Case 2, incall registration follows the scenario shown in Fig. 3. The visited SSP, which receives an incall registration request message from a UPT user sends a message to the FLA through signaling networks. Fig. 4 shows an incall registration scenario for Case 3. If a UPT user requests incall registration to a visited SSP in a remote RA, the visited SSP receives an incall registration request message and sends it to the SCP(H) through signaling networks. The SCP(H) then sends a UPT user's incall registration request message to a new FLA. When the UPT user does not deregister the previous incall registration, the SCP(H) can deregister the UPT user's previous registration. Fig. 5 shows an incall registration scenario in Cases 4 and 5. When a UPT user reregisters from a local RA to a new RA, the visited SSP sends an incall registration message to the FLA through signaling networks. In Case 6, the incall registration scenario follows the procedures shown in Fig. 4.

It is assumed that when a UPT user leaves an LSTP region, the UPT user deregisters a previous incall registration. Fig. 6 shows an explicit incall deregistration scenario where a UPT user who was previously registered on a terminal in the home RA, deregisters at home, local or remote RA. Fig. 7 shows an explicit incall deregistration scenario where a UPT user who was previously registered on a terminal out of the home RA, deregisters at home, local or remote RA. In this case, SSP sends an incall deregistration request message to the FLA through signaling networks and the FLA sends a deregistration message to the previous SCP(V). In the architecture, which uses the FLA, a call is initiated by a UPT user sending a call initiation signal to an SSP. The SSP of the calling UPT user sends a location request message to the SCP(H). The SCP(H) sends a location request message to the FLA of the called UPT user. 4. Mathematical analysis In UPT environments, signaling messages are sent for incall registration/deregistration and call delivery through signaling networks. An example of UPT network architecture is considered, as shown in Fig. 8. Each SSP/SCP(V) is connected to the rest of the network through an LSTP, while a RSTP connects a number of LSTPs. One major function of the LSTP and the RSTP is to route signaling messages. The

E.S. Bae et al. / Computer Communications 23 (2000) 17631776

1767

Fig. 4. Incall registration scenario in Cases 3 and 6.

cost of sending a signaling message through signaling networks may vary depending on the locations of SSPs, SCP(V)s, and SCP(H). The cost of sending a message through a selected type of link is here assumed to be the same regardless of the actual length of the link. To represent the cost of each network element the following notations are made:
q CSCPH : cost for a query to the SCP(H) or an update of the SCP(H); q CSCPV : cost for a query to the SCP(V) or an update of the SCP(V); s CLSTP : cost for routing a signaling message by LSTP; s CRSTP : cost for routing a signaling message by RSTP; s Cla : cost for sending a signaling message through a local A-link; s Cra : cost for sending a signaling message through a remote A-link; s Cd : cost for sending a signaling message through a Dlink; s CFLA : cost for sending a signaling message from the SSP to the FLA;

s CSCPH : cost for sending a signaling message from the SSP to the SCP(H). s s s s The CFLA is the sum of 2 p Cla and CLSTP ; and the CSCPH is s s s s s the sum of CLSTP ; CRSTP ; Cla ; Cra and Cd :

4.1. Location registration cost Let tr and tRA be independent and identically distributed (i.i.d.) random variables representing the incall registration inter-arrival time and the RA residence time of a UPT user, respectively. tr is assumed to be exponentially distributed with a rate of l r. fRA(t) is p the probability density function of tRA and fRA s is the Laplace transform of fRA(t). The mean of tRA is 1=lRA : Fig. 9 shows an embedded Markov chain model, which represents the mobility and incall registration patterns of UPT users. State i is dened as the number of RA crossings since the occurrence of the last incall registration. The state

1768

E.S. Bae et al. / Computer Communications 23 (2000) 17631776

Fig. 5. Incall registration scenario in Cases 4 and 5.

transition probability from state i to state i, aij is given by V b 1 2 r; j i 1 1 b ` j0 aij r; 1 b b X 0; otherwise where r is the probability that one or more than one incall registration occurs between two RA crossings. r is given by 1 p r 1 2 e2lr t fRA t dt 1 2 fRA lr : 2
t0

users reside uniformly in RAs and that the probabilities of UPT user movement from one RA to one of the four neighboring RA are the same. Let PI and PO denote the probabilities that a UPT user is located in an RA with an FLA, and in an RA without an FLA in the LSTP region, respectively. PI and PO are expressed as: PI 1 ; m2 PO 1 2 1 : m2 4

pi is the equilibrium state probability of state i. It is obtained by pi 1 2 ri p0 ; 3 where p0 r: Location information is updated when both registration and deregistration occur with a probability Prd, when only registration occurs with a probability of Pr, or when only deregistration occurs with a probability of Pd. It is assumed that RAs are square and each LSTP region consists of m m RAs. It is also assumed that UPT

Let a kl(n) represent the probability that a UPT user who previously registered at the kth RA moves to the lth RA after n RA crossings. The probability a kl(n) is obtained as:

akl n

f n; x; y ; 4n

where f(n,x,y) is the number of possible paths from the kth RA to the lth RA after n RA crossings. a kl(n) is derived in Appendix A. PIH, PIL, PIR, POH, POL and POR denote the probabilities of location updates of a UPT user, where subscripts denote the following:

E.S. Bae et al. / Computer Communications 23 (2000) 17631776

1769

Fig. 6. Deregistration scenario in the FLA.

I/O: a user's location inside/outside the RA with an FLA at the time instant of the current registration; H/L/R: a user's location at home/local/remote RAs at the time instant of the next registration. These probabilities can be expressed as follows: f n; 0; 0 ; PIH n a11 n 4n a1l n; PIL n
l[A l1

COLr, CORr, CIHd, CILd, CIRd, COHd, COLd, CORd represent the costs for incall registration and deregistration according to the change of user locations where the subscripts of C are classied as follows: I/O: a user's location inside/outside the RA with an FLA at the time instance of the current registration; H/L/R: a user's location at home/local/remote RAs at the time instance of the next registration;
Table 2 The costs for incall registration and deregistration CIHrd CILrd CIRrd COHRd COLrd CORrd CIHr CILr CIRr COHr COLr CORr CIHd CILd CIRd COHd COLd CORd
q 3 p CSCPV q s 4 p CSCPV 1 6 p CFLA q q s 3 p CSCPH 1 3 p CSCPV 1 10 p CSCPH q s 4 p CSCPV 1 2 p CFLA q s 5 p CSCPV 1 8 p CFLA q q s 3 p CSCPH 1 3 p CSCPV 1 10 p CSCPH q 3 p CSCPV q s 4 p CSCPV 1 6 p CFLA q q s 3 p CSCPH 1 2 p CSCPV 1 8 p CSCPH q 3 p CSCPV q s 4 p CSCPV 1 6 p CFLA q q s 3 p CSCPH 1 2 p CSCPV 1 8 p CSCPH q 3 p CSCPV q s 3*CSCPV 1 6 p CFLA q s 3 p CSCPV 1 6 p CFLA q s 4 p CSCPV 1 2 p CFLA q s 4 p CSCPV 1 8 p CFLA q s 4 p CSCPV 1 8 p CFLA

6 7

PIR n 1 2 PIH n 2 PIL n; POH n 1 A ak1 n; m 2 1 k[

2 k1

8 9

POL n

m2

1 akl n; 2 1 k[A l[A

k1 l1

10

POR n 1 2 POH n 2 POL n: A represents a set of RAs in the LSTP region.

11

CIHrd, CILrd, CIRrd, COHrd, COLrd, CORrd, CIHr, CILr, CIRr, COHr,

1770

E.S. Bae et al. / Computer Communications 23 (2000) 17631776

Fig. 7. Deregistration scenario out of the FLA.

r/d/rd: registration only/deregistration only/registration and deregistration. These costs are obtained using the costs s s CFLA ; and CSCPH : For example, when a UPT user previously registered in an FLA registers at a home RA, the home RA is accessed three times, as shown in Fig. 3. q Thus, the cost CIHrd is given by 3 CSCPV : The cost values for incall registration and deregistration according to the change of user locations are given in Table 2. Let CR(i) denote the incall registration cost when a UPT user stays in state i. The incall registration cost is given by: CR i Prd {PIH i p CIHrd 1 PIL i p CILrd 1 PIR i p CIRrd
1POH i p COHrd 1 POL i p COLrd 1 POR ipCORrd }
q CSCPV ; q CSCPH ;

The average incall registration cost per unit time is given by: CR lr C HR ; where C HR
1 k0

13

pk CR k:

4.2. Call delivery cost The previous registration is automatically deregistered when a UPT user makes an incall registration on a terminal. Otherwise, a UPT user may cancel a previous registration before the next registration. It is assumed that users may request deregistration at random points with equal probabilities and that call arrivals follow a Poisson process. It is also assumed that deregistration and call arrivals are mutually independent. The probabilities that a terminating call arrives when a called user is registered or deregistered are denoted by PON or POFF, respectively.

1Pr {PIH i p CIHr 1 PIL i p CILr 1 PIR i p CIRr 1POH i p COHr 1 POL i p COLr 1 POR i p CORr } 1Pd {PIH i p CIHd 1 PIL i p CILd 1 PIR i p CIRd 1POH i*COHd 1 POL i*COLd 1 POR i*CORd }: 12

E.S. Bae et al. / Computer Communications 23 (2000) 17631776

1771

Fig. 8. A reference UPT network architecture.

PON and POFF are given by: 1 y x 1 y x 1 fXuY xfY y dy dxfY y dy PON 0 0 y 0 0 y y 1 1 1 fY y dy ; 2 0 2 POFF 1 2 PON
1 2

where X and Y denote the time intervals from registration to deregistration and to the next registration, respectively. The call delivery cost per unit time CD is then written as: (14)
D D CD {Prd CrD 1 1 2 Prd PON Cr 1 POFF Cd }lc ;

16

15

D where CrD and Cd represent the call delivery cost for the user having registered and deregistered, respectively. l c represents the call arrival rate.

Fig. 9. Embedded Markov chain model.

1772

E.S. Bae et al. / Computer Communications 23 (2000) 17631776

q q D s Cr CSCPH 1 CSCPV 1 4 p CSCPH ; q q D s Cd CSCPH 1 CSCPV 1 4*CSCPH :

5. Numerical examples
IS IS Let CR and CD represent the per unit time incall registration cost and the delivery cost in the IS-41 based scheme, respectively (refer to Appendix B). In order to compare the cost of the FLA scheme with the IS-41 scheme, IS IS IS IS the relative cost CR =CR ; CD =CD ; CR 1 CD =CR 1 CD are introduced. The residence time tRA is assumed to follow an exponential distribution with a mean of 1/l RA. The RMR is dened as lr =lRA : A high RMR indicates that a large number of incall registrations are processed between two consecutive RA crossings. The CRR is also dened as l c/l r. The size of an LSTP region, m pm is set to be 64 m 8 and the value of RMR is varied from 0.5 to 5. The access cost for SCP(H) is higher than for SCP(V) because more signaling links and one more switch are involved in transferring signaling messages to the SCP(H). The cost for accessing SCP(V) is higher than for sending a signaling message from the SSP to the FLA because the processing delay for a signaling messages at the SCP(V) is higher than the routing delay for a signaling message at the LSTP. Four sets of values for cost parameters are considered in Table 3. Fig. 10 shows the relative location update cost for varying the value of RMR when the incall deregistration probability is 0.5. Since the information in the FLA is updated whenever registration occurs within an LSTP region, the location update cost decreases as the value of the RMR increases. The results show that the relative location update cost becomes greater in descending order of the value of CH and CRO. Fig. 11 shows the relative call delivery cost as a function of the incall deregistration probability. The relative call delivery cost decreases as the incall deregistration probability increases because of the decreased database access frequency for determining a user's location. Fig. 12 shows the relative total cost for varying the value of RMR when the CRR and the incall deregistration probability are xed. The results show that the proposed archi-

tecture with the FLA is more efcient than the architecture based on IS-41 for a lower mobility and for a higher access cost to the SCP(H). Fig. 13 shows the relative total cost for varying the value of RMR when the cost parameter of Set 2 is used and incall deregistration probability is 0.5. Since the call delivery cost is higher in the architecture using the FLA than in the IS-41 based architecture, the relative total cost decreases as the RMR increases and the relative total cost is higher as the CRR increases. Fig. 14 shows the relative total cost for varying the incall deregistration probability when the cost parameters of Set 2 are used and the CRR is xed. The architecture with FLA has a lower total cost for a lower mobility, for more frequent incall registration, and for a higher incall deregistration probability.

6. Conclusions An intelligent network architecture using FLA to support UPT services is proposed. In the proposed scheme, when location updates occur within an LSTP region, information related to the location of a UPT user in FLA needs to be updated instead of updating information in both the SCP(V) and the SCP(H). The information in SCP(H) is updated when a UPT user moves from one LSTP region to other LSTP region. Although the call delivery cost is greater in the proposed architecture compared with in the IS-41-based scheme, the proposed architecture is more efcient under the following conditions: with a higher probability of cancelling a previous registration prior to a subsequent registration; with a lower mobility of a UPT user; with a higher probability of registration and deregistration within the same RA; with a higher cost for delivering signaling message to an SCP(H). This study can be extended UPT networks which fully support both personal mobility and terminal mobility. Appendix A. Derivation of the probability a kl(n) The probability that a UPT user originating at the kth RA moves to the lth RA after n RA crossings is derived. Let x, y represent the distance between the two RAs along x- and ydirections such that x uxk 2 xl u y uyk 2 yl u; A1 A2

Table 3 Cost parameters Set 1 2 3 4

q CSCPV q CSCPH s CFLA s CSCPH

3 3 3 3

3 6 9 12

1 1 1 1

2 3 4 5

where (xk, yk) and (xl, yl) are the coordinates for the kth and

E.S. Bae et al. / Computer Communications 23 (2000) 17631776

1773

1.6

1.4 Relative location update cost

Set 1 Set 2 Set 3 Set 4

1.2

0.8

0.6

0.4 0.5

1.5

2 2.5 3 3.5 Ratio of Registration to Mobility

4.5

Fig. 10. Relative location update cost for RMR Pd 0:5:

the lth RAs, respectively. For two-dimensional coverage area with square RAs shown in Fig. 2, a UPT user can move in four directions: `left', `right', `up' and `down'. The number of `right' movement to different RAs is denoted by i0 # i # n: Given that the total number of movements is n, the number of `left', `up' and `down' movements can be expressed in terms of i, x and y. We derive the number of `left', `up' and `down' movements by l(i, x, y), u(i, x, y), d(i, x, y), respectively. li; x; y i 2 x: ui; x; y di; x; y
1 2 1 2

RAs when exactly n movements are performed is given by: f i; x; y

n i0

gi; x; y;

A6

where g(i, x, y) is dened as follows: V b n! ; b b b i!l!u!d! b b ` b b b b b b X

A3 A4 A5 gi; x; y

n i 1 l 1 d 1 u; x 1 y mod 2 n mod 2; n $ x 1 y; l . 0; u . 0; d . 0 otherwise A7

n 2 2i 2 x 1 y: n 2 2i 2 x 2 y:

The number of possible paths from the kth to the lth

2 1.8 Relative call delivery cost 1.6 1.4 1.2 1 0.8 0.6

0;

Set 1 Set 2 Set 3 Set 4

0.2

0.4 0.6 Incall deregistration probability

0.8

Fig. 11. Relative call delivery cost for Pd.

1774

E.S. Bae et al. / Computer Communications 23 (2000) 17631776

1.6

1.4

Set 1 Set 2 Set 3 Set 4

Relative total cost

1.2

0.8

0.6

0.4 0.5

1.5

2 2.5 3 3.5 Ratio of Registration to Mobility

4.5

Fig. 12. Relative total cost for RMR Pd 0:5:

The probability a kl(n) is given by:

akl n

f n; x; y ; 4n

A8

where the numerator represents the number of possible paths from the kth RA to the lth RAs after n movements, and the denominator represents the number of possible path that the UPT user can travel in n movements when the destination is not specied. Appendix B. IN architecture based on IS-41 An IN architecture based on IS-41 needs two types of database: SCP(H) and SCP(V). Fig. B1 shows an IN architecture based on IS-41. This architecture consists of N SSPs, N SCP(V)s and one SCP(H). Each SSP is connected to the corresponding SCP(V) through a direct link, and any SSP
1.6

can access data in an SCP(H) through signaling networks. The conventional functions of the SCP(H) and the SCP(V) are similar to the HLR and VLR in IS-41, respectively. The incall registration scenario based on the IS-41 architecture is the same as the scenario in Fig. 5 except that the SCP(V) sends an in call registration request message to SCP(H) instead of the FLA. The explicit incall deregistration scenario based on IS-41 architecture is the same as the scenario in Fig. 7 except that the visited SSP sends an incall deregistration message to the SCP(H) instead of to the FLA. In the IS-41 based architecture a call is initiated by a UPT user sending a call initiation signal to the SSP. The SSP of the calling UPT user sends a location request message to the SCP(H) and the SCP(H) sends the location of a called user to the SCP(V). The cost in the IS-41 based architecture is dened as follows:

1.4

CRR=0.5 CRR=1.0 CRR=1.5 CRR=2.0

Relative total cost

1.2

0.8

0.6

0.4 0.5

1.5

2 2.5 3 3.5 Ratio of Registration to Mobility

4.5

Fig. 13. Relative total cost for CRR Pd 0:5; Set2:

E.S. Bae et al. / Computer Communications 23 (2000) 17631776

1775

1.6

1.4

Pd= 0.2 Pd= 0.5 Pd= 0.8

Relative total cost

1.2

0.8

0.6

0.4 0.5

1.5

2 2.5 3 3.5 Ratio of Registration to Mobility

4.5

Fig. 14. Relative total cost for incall deregistration probability.

Fig. B1. Intelligent network architecture based on IS-41.

1776

E.S. Bae et al. / Computer Communications 23 (2000) 17631776 [4] Recommendation F.851, Universal Personal Telecommunication (UPT)-service description (service set 1), January 1994. [5] Y. Kawanami, Migration from personal mobility to UPT in mobile environment, Proc. of the IEEE Intelligent Network Workshop (IN'95), vol. 1, 1995. [6] M. Campana, P. Chauvaud, H. Gilbert, Security architecture for the UPT service, Proc. of the ISS, vol. 1, 1995. [7] H. Kikuta, M. Fujioka, Mechanism of UPT OutCall Registration over Multiple IN-based networks, Proc. of the IEEE Intelligent Network Workshop (IN'95), vol. 1, 1995. [8] J. Audestad, B. Jacobsen, Universal personal telecommunication and intelligent network architecture, Proc. of the International Conference on Intelligence in Networks (ICIN), vol. 1, 1995, pp. 8994. [9] C. Low, D. Skov, N. Raguideau, WebINan architecture for fast deployment of IN-based personal services, Proc. of the IEEE Intelligent Network Workshop IN'96, vol. 1, 1996. [10] M. Fujioka, S. Sakai, H. Yagi, Hierarchical and distributed information handling for UPT, IEEE Network 4 (6) (1990) 5060. [11] J. Dan, C. Vernhes, IN as a platform for UPT: constraints and requirements, Proc. of the International Conference on Intelligence in Networks (ICIN), vol. 1, 1995, pp. 7982. [12] M. Sullivan, Universal personal telecommunication networking issues, Proc. of the International Conference on Intelligence in Networks (ICIN), vol. 1, 1995, pp. 195199. [13] G.S. Lauer, IN Architectures for Implementing Universal Personal Telecommunications, IEEE Networks March/April (1994) 616. [14] A. Folkestad, Impact of UPT service on network performance and call set-up times: distribution of service logic in SS7, Proc. of the ISS, vol. 2, 1995, pp. 97101. [15] M. Kwiatkowski, Performance modeling of UPT networks, Proc. of the IEEE International Conference on Universal Personal Communications, vol. 1, 1995, pp. 543547. [16] M.Y. Chung, D.K. Sung, Performance analysis of a prole management scheme for incall registration/deregistration in wireline UPT networks, IEICE Trans. Commun. E82-B (5) (1999). [17] M.Y. Chung, S.H. Kang, D.K. Sung, Performance analysis of two schemes for managing information related to incall registration in wireline UPT networks, Computer Communications 22 (1999) 15921606. [18] J.S.M. Ho, I.F. Akyildiz, Local anchor scheme for reducing signaling costs in personal communications network, IEEE Transaction on Networking 4 (5) (1996) 709723. [19] EIA/TIA, Cellular Radio-Telecommunications Intersystem Operations, Technical Report IS-41 Revision B, EIA/TIA, 1991.

IS Crd : cost of a UPT user's incall registration and deregistration; IS Cr : cost of a UPT user's incall registration; IS Cd : cost of a UPT user's deregistration; IS=D Cr : call delivery cost in registration; IS=D Cd : call delivery cost in deregistration; IS CR : incall registration cost per unit time; IS CD call delivery cost.

The cost for sending a signaling message in terms of the cost parameters in the IS-41 based architecture is expressed as
q q IS s Crd 3 p CSCPH 1 2 p CSCPH 1 8 p CSCPH ; q q IS s Cr 3 p CSCPH 1 CSCPH 1 6 p CSCPH ; q q IS s Cd 3 p CSCPH 1 CSCPH 1 8 p CSCPH ; q IS=D s Cr CSCPH 1 2 p CSCPH ; q IS=D s Cd CSCPH 1 2 p CSCPH :

The incall registration cost per unit time and the call delivery cost in the IS-41 based network can be written as
IS IS IS CR Prd Crd 1 Pr CrIS 1 Pd Cd lr ;

B1

IS IS=D IS=D IS=D CD {Prd Cr 1 1 2 Prd PON Cr 1 POFF Cd }lc : B2

References
[1] ITU-T Recommendation F.850, Principles of universal personal telecommunication (UPT), March 1993. [2] ITU-T Draft Recommendation I.373, Network capabilities to support universal personal telecommunication (UPT), July, 1992. [3] ITU-T Draft Recommendation Q.76, Service procedures for universal personal telecommunication. Functional modelling and information ows, 1993.